GRACE - LogiQA Categorization: leaderboard
Metric: Four-class error-categorization macro-F1 (%) on LogiQA inference traces; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 78.77 |
| 2 | Gemini 3 Flash | 59.17 |
| 3 | GPT-5.4 | 52.18 |
| 4 | Gemma 4 31B | 52.02 |
| 5 | Qwen 3.5 27B | 44.27 |
| 6 | Qwen 3.5 35B A3B | 37.04 |
| 7 | GPT-5.4 Mini | 33.71 |
| 8 | Llama 3.1 8B | 32.05 |
| 9 | Qwen 3 14B | 25.19 |
| 10 | Qwen 3 8B | 16.43 |
Interactive version: theaggregate.ai/benchmark?slug=grace-logiqa-categorization · How It Works · Data refreshed daily, snapshot 2026-09-29.